Why It Stopped Being Automatic

A woman works it out somewhere around forty, usually while filling in a form or planning something, and the realisation is oddly specific: she has not seen anybody who is only her friend — not a colleague, not another mother, not a relative — in about four years. Not because of a falling-out. Because there was never an afternoon on which it would have happened.

The conclusion almost everybody draws from that is about themselves: that they are bad at this, or that they became boring, or that something in them closed. This page argues that the conclusion is wrong in a way that is both demonstrable and immediately useful, because what actually changed was not her and not them. It was the mechanism, and the mechanism is not a mystery.

Friendship in the first half of a life is supplied by an environment rather than by effort, and the environment has three specific ingredients. Proximity: the same people are physically near you. Repetition: they are near you again tomorrow, without anybody arranging it. And a setting relaxed enough for a guard to come down — not a meeting, not a transaction, but time with nothing particular being achieved.

A classroom supplies all three for twelve years. A university supplies them for four. A first office floor supplies a weaker version for a while. Then adult life removes all three at once and does not announce that it has: you commute, you work with people you did not choose toward ends that are not yours, and you go home. There is no longer any setting in which the same unrelated people are near you repeatedly with nothing to achieve.

So the friends who survive into a woman’s forties are almost never the ones she liked most. They are the ones where something replaced the mechanism — a standing arrangement, a group that meets, a shared obligation, an annual thing nobody questions. Everybody else drifted, on both sides, symmetrically, and no decision was made by anyone. That is worth holding onto, because the private version of this story is always that they drifted from her.

There is research that puts an actual figure on it, and the figure does more work than any amount of encouragement. Hall, publishing in 2019, estimated the hours of time together required to move between stages of closeness: roughly fifty hours to reach casual friend, around ninety to reach what people call a friend, and more than two hundred before somebody is a close friend.

Read those against an ordinary adult year and the arithmetic becomes obvious rather than tragic. Lunch once a month with somebody new is about twelve hours a year, which arrives at casual friend in something over four years. A weekly class of ninety minutes is about seventy hours a year and crosses the second threshold inside eighteen months. The difference is not warmth or effort or personality. It is frequency, and frequency is a thing on a calendar.

Which reframes the sentence a great many women say to themselves. I cannot make friends any more is a claim about a person. I have not spent fifty hours with anybody new since before the pandemic is a claim about a calendar, and only one of those two has anything you can do about it on a Tuesday.

A great deal of avoidable pain in this country comes from one category error, and naming it removes most of it. ママ友 is a functional relationship. It is organised around children, it exists because two women are in the same school district in the same years, and it has real warmth in it — and it ends, usually completely, when the children’s schools diverge. That ending is not a betrayal and was not caused by anything either woman did. It is what happens when the thing the relationship was organised around is removed.

The same is true of the colleague, with a ritual attached that makes it unusually clear: the leaving party exists precisely because everyone understands the relationship is bounded by the building. Which is why leaving a job in this country can remove an entire social world in a single afternoon, and why women who left work to have children so often describe the loss as though something was done to them. Nothing was. The container was closed and everything in it went with it.

And then 疎遠, which is the distinctively difficult one because there is nothing in it to resolve. No argument, no grievance, no moment. Two people simply stopped, over years, for no reason at all — and the barrier to picking it up again is precisely the absence of a reason. If there had been a falling-out there would be something to apologise for and a script for doing it. Instead there is only the elapsed time, and the elapsed time is the whole of the obstacle.

The barrier to re-contacting somebody after years is a prediction, and the prediction has been tested. She expects it to be awkward, to require explaining the gap, and to be received as slightly strange — and the other person, she assumes, has moved on and will find it intrusive.

A set of thirteen preregistered experiments involving around six thousand participants, published in 2022, put people on both sides of exactly that situation and compared what the initiators predicted with what the recipients reported. Initiators consistently and substantially underestimated how much the contact was appreciated, and the effect ran through surprise: the more unexpected the message, the more it was valued. The asymmetry is the finding. She is running an accurate model of her own awkwardness and an inaccurate model of their gladness.

The limit belongs with it: that is an effect about appreciation measured across many people, not a guarantee about one particular old friend on one particular Tuesday. What it does establish is that the reason she has not sent the message is a forecast, that the forecast is systematically wrong in a known direction, and that the message does not have to explain the gap. Nobody on the receiving end has been keeping score of the years. They are almost always just pleased to be thought of.

If the environment was what supplied friends, then the thing to build is an environment rather than a friendship. That sounds like an evasion and it is the opposite: it is the only version of this that has ever worked for anybody, and it explains why the usual advice fails.

The usual advice is an event: a party, a one-off class, a gathering somebody invited her to. Events fail because they supply proximity once and repetition never, and one evening with twenty strangers produces about three hours against a fifty-hour threshold. A recurring thing with the same eight people produces the same three hours and then does it again in a week, and the eight people are still there in November.

So the specification is dull and exact. Recurring rather than one-off. The same people rather than a rotating crowd. Low enough stakes that missing one week is nothing. And — this is the ingredient that gets designed out most often — some purpose other than friendship, because two adults who have met explicitly to become friends are both performing, and performing is the condition under which the guard does not come down. A choir, a class, a volunteer shift, a walking group, a regular study session. The activity is a pretext and it is supposed to be one.

And expect nothing for a long time, because the numbers say to. Eighteen months of a weekly thing is the realistic shape of this, and a woman who joins something in April and concludes in June that it did not work has stopped at about eleven hours. The most common failure here is not choosing badly. It is quitting inside the threshold.

A different shortage, which is often mistaken for the first one and has a different remedy: she is not short of people. She has three friends she has known since university, they meet twice a year, and the conversation is entirely logistics, children, parents and health. She could not tell any of them that she has been crying in the car, and she does not know when that became true.

The mechanism here is not distance, it is staleness of information. A friendship formed at twenty-two was built on a shared situation, and both people then changed for twenty years without updating each other — because twice a year is enough to maintain contact and nowhere near enough to transmit change. So each woman is talking to a version of the other from around 2009, carefully, because the version she is addressing would not understand the thing she wants to say. Both are doing this. Neither knows the other is.

The move is not to find new friends and it is not to have a conversation about the friendship, which is excruciating and rarely works. It is to say one true thing, small and unannounced, in the middle of an ordinary exchange — one sentence about something actually going on, offered without a preamble and without requiring a response in kind. That is a test of the channel rather than a demand on it, and it costs almost nothing to run.

Frequently the other woman answers with something equally true and slightly too fast, which is what somebody sounds like when they have been waiting. Sometimes she changes the subject, and that is information rather than a verdict — some friendships are genuinely containers for a shared past and nothing else, and that is a real thing to have rather than a failed version of something better. What is not worth doing is spending another decade addressing a person from 2009 and concluding that nobody knows you.

Some women are not describing a scheduling problem. They find this genuinely difficult, always have, and have a lifetime of evidence that they are worse at it than other people — and being told the mechanism is environmental can read as a polite refusal to believe them.

So it is worth saying that the argument above helps her most rather than least, and specifically why. The two things that make this hard — having to perform on first meeting, and having to generate conversation from nothing — are both features of the event format, which is the format that fails anyway. A recurring low-stakes setting with a purpose removes both: there is something to do, so nobody has to be interesting; familiarity accumulates without anybody deciding it should; and the thing that eventually makes two people friends is having been in the same room forty times, which requires no charm at all.

Do not aim at a best friend. The single-confidante model is the one almost everybody has in mind — one person who knows everything, available at any hour — and it is both rare in adulthood and a poor target to organise around, because it cannot be built directly and failing to produce it reads as total failure. What actually makes a week feel populated is usually four or five people at various middle distances, none of whom knows everything. Aiming at the rare thing is how women end up with none of the ordinary ones.

And do not turn it into a project with a target, which is the trap this library keeps finding in capable women. A friendship that is being managed toward an outcome is a performance, and performing is the exact condition the third ingredient rules out. The same test applies here as everywhere else in this library: if the arrangement does not let you skip a week when you are tired, without a reason and without anybody being disappointed, it is not supplying the thing you went for.

This is a page about being short of company, published by a business that sells company by the hour, so it should say the expensive thing plainly rather than let it be inferred.

A paid hour is not a friendship, does not become one, and should not be bought as a substitute for one. That is not modesty; it is the same structural point this library made about the whole women-facing market. The arrangements that hurt people are the ones that are dishonest about being arrangements — where regard is presented as personal and particular and is in fact contingent on a booking. We are an arrangement. Ours is honest about it. It also cannot do what fifty hours in a choir does, and it is not trying to.

What it honestly is: an hour of undivided attention with a guaranteed ceiling and no accumulated history, which is a real good and a different good. If what is actually missing is four people at middle distance who will still be there in November, an hour here will feel excellent on the day and will not touch it. We would rather say that and lose the booking than take it and have you find out in March.

So: nothing closed in you. An environment that supplied proximity, repetition and a relaxed setting for the first twenty-five years of your life was withdrawn without notice, and the friends who survived it are the ones where something replaced it rather than the ones you liked best. Fifty hours, ninety, two hundred. A weekly thing with the same people crosses the second of those inside a year and a half, and a monthly lunch does not.

And the message you have not sent to somebody you have not spoken to since 2019 does not have to account for the six years. The person at the other end is not keeping the score you think they are keeping, and the research says fairly clearly that you are worse at predicting their gladness than you are at predicting your own embarrassment.